We examined the effects of measurement error on epidemiologic studies. If errors are correlated with the true exposure, as for example when obese males tend to underestimate their caloric intakes, then classical methods to correct for measurement error can lead to exaggerated estimates of exposure effects. A simulation study showed that typical case-control investigations of lung cancer risk from radon exposure in the home have low power because radon exposure measurements are subject to error and because residential mobility reduces differences in exposure. We studied the variability of laboratory measurements of estrogen and its metabolites in serum and urine. These studies indicate that assay variability is usually small compared to variation among women; large sample sizes will be needed, however, to detect modest differences in hormone levels between women with cancer and controls. We showed that the Wald test is unreliable for testing hypotheses concerning components of variance in these studies. We published theory and guidelines to determine how much information subjects with error-prone exposure measurements can add to a study that also includes subjects with "gold standard" measurements. We reviewed the fundamental contributions of Jerome Cornfield and Bradford Hill to the theory of case-control studies as well as the contributions of David Byar. We also discussed the limitations of observational data for drawing inferences on treatment effects and offered a critique of such analyses of treatments for HIV/AIDS. We developed robust permutational methods to analyze community intervention trials, and we developed sequential boundaries for monitoring Phase II trials. We developed a method to select a parsimonious set of questions from a dietary questionnaire that yields similar estimates of nutrient intake to those obtained from the full questionnaire. We developed methods of inference to estimate agreement between two types of exposure measurements that have been categorized based on marginal quantiles, as is done in nutritional studies.